from long to float for regression

HumorDataset.py

@@ -5,6 +5,40 @@ import torch
 import numpy as np
 from nltk.tokenize import word_tokenize
 
+class TextRegDataset(torch.utils.data.Dataset):
+    def __init__(self, texts, labels, word_index, max_len=50):
+
+        self.original_indices = labels.index.to_list()
+
+        self.texts = texts.reset_index(drop=True)
+        self.labels = labels.reset_index(drop=True)
+        self.word_index = word_index
+        self.max_len = max_len
+
+    def __len__(self):
+        return len(self.texts)
+
+    def __getitem__(self, idx):
+        texts = self.texts[idx]
+        tokens = word_tokenize(texts.lower())
+
+        label = self.labels[idx]
+        
+        # Tokenize and convert to indices
+        input_ids = [self.word_index.get(word, self.word_index['<UNK>']) for word in tokens]
+
+        # Pad or truncate to max_len
+        if len(input_ids) < self.max_len:
+            input_ids += [self.word_index['<PAD>']] * (self.max_len - len(input_ids))
+        else:
+            input_ids = input_ids[:self.max_len]
+        
+        # Convert to PyTorch tensors
+        input_ids = torch.tensor(input_ids, dtype=torch.long)
+        label = torch.tensor(label, dtype=torch.float)
+        
+        return input_ids, label
+    
 class TextDataset(torch.utils.data.Dataset):
     def __init__(self, texts, labels, word_index, max_len=50):
 

TEST_CNN_2.py

@@ -5,7 +5,7 @@ from torch.utils.data import DataLoader
 from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
 from tqdm import tqdm  # Fortschrittsbalken-Bibliothek
 from dataset_generator import create_embedding_matrix, split_data
-from HumorDataset import TextDataset
+from HumorDataset import TextRegDataset
 import numpy as np
 import pandas as pd
 import os
@@ -20,7 +20,7 @@ params = {
     "learning_rate": 0.001,
     "epochs": 25,
     "glove_path": 'data/glove.6B.100d.txt',  # Pfad zu GloVe
-    "max_len": 50,
+    "max_len": 280,
     "test_size": 0.1,
     "val_size": 0.1,
     "patience": 5,
@@ -171,9 +171,9 @@ visualize_data_distribution(y)
 data_split = split_data(X, y, test_size=params["test_size"], val_size=params["val_size"])
 
 # Dataset und DataLoader
-train_dataset = TextDataset(data_split['train']['X'], data_split['train']['y'], word_index, max_len=params["max_len"])
-val_dataset = TextDataset(data_split['val']['X'], data_split['val']['y'], word_index, max_len=params["max_len"])
-test_dataset = TextDataset(data_split['test']['X'], data_split['test']['y'], word_index, max_len=params["max_len"])
+train_dataset = TextRegDataset(data_split['train']['X'], data_split['train']['y'], word_index, max_len=params["max_len"])
+val_dataset = TextRegDataset(data_split['val']['X'], data_split['val']['y'], word_index, max_len=params["max_len"])
+test_dataset = TextRegDataset(data_split['test']['X'], data_split['test']['y'], word_index, max_len=params["max_len"])
 
 train_loader = DataLoader(train_dataset, batch_size=params["batch_size"], shuffle=True)
 val_loader = DataLoader(val_dataset, batch_size=params["batch_size"], shuffle=False)
@@ -187,7 +187,10 @@ model = EnhancedCNNRegressor(
     num_filters=params["num_filters"],
     embedding_matrix=embedding_matrix,
     dropout=params["dropout"]
-).to(device)
+)
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = model.to(device)
 
 criterion = nn.MSELoss()
 optimizer = optim.Adam(model.parameters(), lr=params["learning_rate"], weight_decay=params["weight_decay"])
@@ -340,3 +343,15 @@ test_rmse = np.sqrt(mean_squared_error(test_labels, test_preds))
 test_mae = mean_absolute_error(test_labels, test_preds)
 test_r2 = r2_score(test_labels, test_preds)
 print(f"Test RMSE: {test_rmse:.4f}, Test MAE: {test_mae:.4f}, Test R²: {test_r2:.4f}")
+
+
+# plot distribution of predicted values and true values
+plt.figure(figsize=(10, 6))
+plt.hist(test_labels, bins=20, color='skyblue', edgecolor='black', alpha=0.7, label='True Values')
+plt.hist(test_preds, bins=20, color='salmon', edgecolor='black', alpha=0.7, label='Predicted Values')
+plt.title('Distribution of Predicted and True Values')
+plt.xlabel('Score')
+plt.ylabel('Frequency')
+plt.legend()
+plt.grid(axis='y', linestyle='--', alpha=0.7)
+plt.show()